Identification of Network Disruptions Using the Fuzzy K-Nearest Neighbor Algorithm in Case Based Reasoning at STMIK Pranata Indonesia

Case-Based Reasoning (CBR) is a computer reasoning system that uses old knowledge to solve new problems. CBR provides solutions to new cases by looking at old cases that are closest to new cases. This would be very beneficial as it would eliminate the need to extract models as required by rule-based systems. The main problem in this research is the frequent disruption/difficulty for students and staff to connect to the network at the Indonesian institutional system. The method used is fuzzy K-Nearest Neighbor (FK-NN) is a variant of the K-Nearest Neighbor (KNN) method with fuzzy techniques. Using case based reasoning to make it easier for the system to reason, which prioritizes the use of information and previous knowledge to represent experience as a source of learning basis for reasoning. This system can help the IT team at STMIK Pranata Indonesia to more quickly handle types of network disturbances. Testing the system using a user acceptance test produces acceptable results and is used properly according to user needs. The scoring method uses a Likert scale with acceptance results of 94% with the criteria of strongly agree. The highest accuracy was obtained from test results with K = 5 with a value of 94.33%


INTRODUCTION
The development of science and technology is currently very rapid, commensurate with increasingly complex human problems.Humans have limitations in thinking and remembering, it takes several seconds, minutes or even hours to remember events that occurred in the past.Therefore, we need a tool or system that can help the work processes carried out by humans.A computer is a tool used to process data according to procedures that have been formulated.Computers can be used to assist the work processes carried out by humans.Some work processes that were previously carried out manually by humans can be carried out automatically by computers.
Identifying problems in the network certainly takes time, the process of finding this problem becomes a new problem.Speed and accuracy are needed to solve this problem.This process certainly requires experience and flying hours as well as strong analysis in identifying problems in the network so that solutions can be quickly found.A computer network system is a system consisting of computers and other network devices that work together to achieve the same goal (MADCOM, 2019).
Case based reasoning (CBR) adalah jenis metode yang using information or knowledge from historical cases to solve problems in similar cases.It allows people to gain knowledge through the study of historical cases, so as to match similar historical cases from a database, and modify and improve these historical cases accordingly, so as to draw conclusions about current problems.CBR is an expert system methodology that can be used in the field of network telecommunications which is appropriate to be applied to overcome communication network obstacles, widely used in fields including: law, medicine, computational engineering, manufacturing, scheduling, environment, finance, technology and others. .Syofyan and Kridiagung (2017) prove that CBR can diagnose symptoms of damage to two-wheeled vehicle engines using the retrieve, reuse, revise and retain cycle by calculating problems that have already occurred so as to provide the right solution regarding definite damage based on selecting damage symptoms.selected with predetermined weighting.Nugraha et al. (2017) states that the average accuracy is 82% and 71% in the Fuzyy K-Nearest Neighbor (F-KNN) and K-Nearest Neighbor (KNN) methods.F-KNN calculates the membership degree value in each class which plays a role in prediction.class, so F-KNN is better because it doesn't just make decisions based on the majority number that appears in KNN.In the research that will be carried out, case based reasoning is used to diagnose problems in the network, while the fuzzy k-nearest neighbor algorithm method is applied to support the search for solutions based on the closeness of values from previous data.
The wide coverage of the LAN (Local Area Network) makes it difficult to find damage or errors that cause disruption to the LAN.Therefore, we need a technology that allows humans to know the causes of disturbances that occur in the network.An expert system is a computer-based advisory program that tries to imitate the thinking process and knowledge of an expert in solving specific problems.With an expert system, someone can analyze based on seeing the results of the diagnosis of symptoms or problems that occur on the network.STMIK Pranata Indonesia, which is located in Bekasi City, often experiences network disruptions which result in work being hampered so that a lot of time is wasted and limited staff in the IT Infrastructure section is a problem.
Based on the problems mentioned above, a network disturbance identification system is needed that is able to provide alternative solutions to problems that occur and provide solutions to network disturbances that users experience.Apart from that, this system will also be able to provide learning to users who are unfamiliar with network disturbances by reading solutions to problems.Users can learn about network disturbances experienced by practicing the solutions provided by the system while waiting for IT to handle the disturbances that occur.

LITERATURE REVIEW A. Fuzzy Logic
Fuzzy logic was first put forward by an Iranian national who was a professor at the University of California at Berkeley named Lotfi Zadeh in 1965.Fuzzy logic is knowledge that has the ability to bridge measurable machine language with human language which tends to be immeasurable.Fuzzy logic has uncertain values or ambiguity between true and false, a value can be true and false at the same time, but how true and false a value is depends on the membership weight it has.With fuzzy logic, human language is implemented into language easily and efficiently.Fuzzy logic is an appropriate way to map an input into an output space.
A Crisp set is defined by the items in that set.If a is a member of A, then the value associated with a is 1.However, if a is not a member of A, then the value associated with a is 0. The notation A = {x P(x)} indicates that A contains item x with P(x) is true.If XA is a function of characteristic A and property P, then it can be said that P(x) is true, if and only if XA(x) = 1.
Fuzzy sets are based on the idea of expanding the range of characteristic functions such that the function will cover real numbers in the interval [0,1].The membership value indicates that an item in the universe of discussion is not only at 0 or 1, but also at values that lie in between.In other words, the truth value of an item is not simply true or false.A value of 0 indicates false, a value of 1 indicates true and there are still values that lie between true and false.
Fuzzy sets can be represented by membership functions.Membership functions can take many forms, but there are some common examples that arise in real applications.The membership function can be selected by the user according to their experience or can be decided by machine learning methods, and can be depicted as a curve, where the x-axis represents the value of the universe and the y-axis represents the degree of membership.The membership function of a strict set must be like a clock signal because the y-axis value is 0 or 1; The membership function of a fuzzy set may be triangular, trapezoidal and so on (Jing Lu, 2016).
There are several operations defined to modify or combine fuzzy sets.The membership value of the results of the operation of two sets is called the free strength (α-predicate).The 3 basic operators created by Zadeh as the basic operators of fuzzy logic are: 1. Operator AND The α-predicate of two sets modified with the AND operator is the smallest membership value between the elements in the sets concerned.μA∩B = min ( μA(X), μB(y)).2. Operator OR The α-predicate of two sets modified with the OR operator is the largest membership value between elements in the sets concerned.μA∩B = max ( μA(X), μB(y)).

Operator NOT
The α-predicate of two sets modified with the NOT operator is the membership value of the set minus 1. μA'=1-μA (x)

B. K-Nearest Neighbor
The K-Nearest Neighbor Algorithm is an algorithm that performs classification based on the proximity of the location (distance) of data to other data.The working principle of K-Nearest Neighbor itself is to find the shortest distance between the data being evaluated and the K nearest neighbors in the data.Before finding the distance between data evaluated in the K-NN algorithm, preprocessing or normalization must be carried out first.Preprocessing itself aims to obtain standard values for all attributes or indicators in the calculation.Salamun (2017) The formula for calculating similarity weights with nearest neighbors is as follows: Similarity (problem,case) = S1* W1 + W2 + …+ Sn*Wn W1+ W2 +.+Wn (2.1) Information : S = Similarity (similarity value where Similar = 1, not similar = 0) W = Weight (given weight) Weighting is determined based on learning results or observations in cases.The more influence a symptom has on a case, the higher the weight, and vice versa.

C. Fuzzy K-Nearest Neighbor
Fuzzy K-Nearest Neighbor is a classification method that combines fuzzy techniques with the K-nearest neighbor classifier.The fuzzy K-nearest Neighbor algorithm assigns class membership values to test data rather than placing test data in a particular class.FK-NN is a classification method used to predict test data in each class, then the class with the largest membership degree value from the test data is taken as the predicted class.The advantage is that the membership values in the test data class provide a level of guarantee for the classification results.
For example, if a vector is given a membership value of 0.9 in the first class and 0.1 in the other two classes, the researcher can be quite sure that the class with a membership value of 0.9 is the class that the vector belongs to.On the other hand, if a vector is given a membership value of 0.55 in the first class, 0.44 in the second class and 0.01 in the third class.So researchers should be hesitant to assign vectors based on these results.However, it can be believed that the vector does not belong to the third class.In cases like this, it needs to be examined further to determine the classification, because it has a high degree of membership in two classes, namely one and two.Giving membership values by this algorithm is clearly useful in the classification process.
The normalization process that is commonly used is the Min-Max normalization calculation using equation (2.2) as below.
V` = v(x) -min(x) Range(x) V` = Valued between 0 and 1 V(x) = The attribute value to be normalized Max(x) = Maximum value of a parameter Min(x) = Minimum value of a parameter Range(x) = Value of (max(x)-min(x)) Preprocessing itself aims to obtain standard values for all attributes or indicators in the calculation.The Euclidean distance calculation is used to calculate the closeness distance between the test data and the training data.The Euclidean distance calculation is as in equation 2.3 below.

2
(2. 3) The basis of the FKNN algorithm is to determine the membership value as a function of the distance vector from the KNN.Before calculating the FKNN membership value, the process of equation ( 2 Case based reasoning (CBR) is the main paradigm in automatic reasoning and machine learning.In CBR, someone who does reasoning can solve a new problem by paying attention to similarities with one or several solutions to previous problems.in designing an expert system by making a decision from a new case which is carried out based on the solution from the previous case.The concept contained in this method is to find an idea that will be used to carry out documented experience in solving a new problem.
Case based reasoning is a reasoning method in an expert system, the reasoning used is case based reasoning.Case based reasoning uses previous experience in similar cases to understand and solve new problems.The way CBR works is by comparing new cases with old cases.If there are similarities between the new case and the old case, then CBR will provide answers to the old case for the new case.If there is no match, CBR will adapt by entering the new case into the case-based database, thus adding new knowledge to CBR.According to (Main et al, 2001) in research (Mulyana & Hartati, 2009) Case based reasoning is defined as a methodology for solving problems by utilizing experience.Case Based reasoning consists of four main steps, namely: 1. Retrive Retrive that is, taking the same problem again.In this step, a search or calculation process is carried out for cases that have similarities.

Reuse
Reuse that is, reusing the information and knowledge in the case to solve new problems.In this step, solutions are sought from similar cases in previous conditions for new problems.

Revise
Revise namely reviewing the solutions provided.In this step, solutions are sought from similar cases in previous conditions for problems that occur later.

Retain
Retain is to integrate/save new cases that have succeeded in getting a solution so that they can be used by subsequent cases that are similar to that case.However, if the solution fails, then explain the failure, and provide the solution that was used and test it again.
The process in CBR can use various techniques, including the Fuzzy K-Nearest Neighbor algorithm used in the CBR process to carry out a classification process for objects based on learning data that is closest to the object, or it can be said to calculate the level of similarity (distance) of a case to another case.others based on attributes that are defined based on certain weightings and then the level of similarity (distance) of all attributes will be carried out.The relationship between these steps can be presented in Figure 1, Figure 1.Life Cycle CBR (Aamodt and Plaza, 1994) In education, case based reasoning expands a person's knowledge by using new experiences into memory/databases that are used in solving problems in the future.Case based reasoning is also heavily influenced by research results in the field of cognitive science.The process in case-based reasoning is close to and like the reflection of reasoning in humans, where when faced with the same situation the problem is solved by humans in the same way as the solution in CBR.Just like humans are able to reason, case based reasoning was developed to reason like humans, through case based reasoning you can match and retrieve solutions from the past which are stored and used to solve current problems.

E. Pre hypertext Prepocessor (PHP)
According to Rismon H. Sianipar ( 2019) PHP is about learning: how to do different things, how to improve basic techniques, and how PHP programming techniques have a strong PBO foundation.PHP is a programming language that is widely used to handle the creation and development of websites and can be used in conjunction with HTML.PHP was first created by Rasmus Lerdorf in 1994.Initially PHP was an abbreviation for "Personal Home Page Tools".Later it was changed to FI ("Forms Interpreter").Since version 3.0, the name of the language was changed to "PHP: Hypertext Prepocessor" with the abbreviation "PHP".The latest version of PHP is version 8. Based on a survey https://w3techs.com/ in January 2022, more than twenty million websites use PHP, including Wikipedia, Facebook, and Instagram.
When called from a web browser, a program written in PHP will be parsed on the web server by a PHP interpreter and translated into an HTML document, which will then be displayed back to the web browser.Because PHP program processing is carried out in a web server environment, PHP is said to be a serverside language.Therefore, as previously stated, the PHP code will not be visible when the user selects the "View Source" command on the web browser they are using.Apart from using PHP, web applications can also be built with Java (JSP -JavaServer Pages and Servlet), Perl, or ASP (Active Server Pages).

F. User Acceptance Test (UAT)
According to Betha (2006) in research (Wahyuningsih and Wibawa, 2017) User Acceptance Test (UAT) or user acceptance test is a testing process by users which is intended to produce documents which become evidence that the software that has been developed can be accepted by users, if the results testing can be considered to meet the needs of the user.UAT objectives: a) Testing the system to see whether it meets what is required b) Provide confidence that the system that has been developed meets the requirements c) As a complement to a number of approved requests Benefits of UAT: a) Increase client confidence about the software's potential to meet requirements b) Through defect identification, testing ensures that the software is stable and in a workable condition c) Client satisfaction increases, because they are more confident that the system meets requirements d) Obtain a system that complies with the system's functional specifications

METHODOLOGY
The methodology used is a descriptive quantitative method which aims to obtain a more accurate and complete picture of the object under study.Data collection techniques were carried out using observation, documentation and literature study techniques.Quantitative methods are a form of research that is carried out systematically, regularly and in detail.In practice, this research method focuses on the use of numbers, tables, graphs and diagrams to display the results of the data/information obtained.

Data
The data taken is in the form of network disturbance symptoms in   The system testing technique that the author uses is the User Acceptance Test (UAT) method to test the application prototype that will be used.This test will be carried out after the prototype has been created, then several respondents will be determined in testing this system.The results of this test are to determine whether the CBR prototype that has been made meets the needs or is still far from sufficient.

A. Case Based Reasoning Calculations
The retrieve process is a process for finding similarity values for new cases with existing cases in the knowledge database.To be able to find similarity values, use the K-nearest neighbor (KNN) algorithm.
Table 2. Default Router Similarity Value Euclidean calculation by calculating the initial data for each record from the testing data minus the record data from the test data, raised to the power of 2, the sum of each attribute and then squared root.Using equation 2.3 as in this Euclidean calculation.The results are obtained after sorting.
Table 12.Euclidean Distance Ordering After getting the k value determined, a fuzzy process is carried out with the aim of finding the membership value for each class j (agree/disagree category) using equation (2.4).Previously, first find the value of n.Is known :  Testing was carried out using different K values, namely K=10, K=15, K=20, K=25 with the same amount of data, based on the results of the testing experiment, the effect of the K value on accuracy can be seen in Figure 4.5 Testing with a value of K=10 produces an accuracy of 95%.Testing with a value of K=15 produces an accuracy of 93.33%.Testing with a value of K=20 produces an accuracy of 93.33%.Testing with a value of K=25 produces an accuracy of 91.67%.The total average accuracy in this test was 94.33%.
The number of members of class j in a training data n n = The amount of training data used j = Data class Next, calculate the membership value of each class with equation (2.5) membership values on test examples (x, xj) k = nearest neighbor value j = test data membership data variables m = large rank weight m > 1 D. Case Based Reasoning (CBR)

A
. System Design This use case diagram depicts the user selecting the symptoms experienced on a network that is experiencing interference, and the user sees the results of the interference and the results of the network disruption solution.

Figure
Figure 1.System Design B. System Testing TechniquesThe system testing technique that the author uses is the User Acceptance Test (UAT) method to test the application prototype that will be used.This test will be carried out after the prototype has been created, then several respondents will be determined in testing this system.The results of this test are to determine whether the CBR prototype that has been made meets the needs or is still far from sufficient.

Figure
Figure 2. Membership Value

Figure 5 .
Figure 5.Test Results Effect of K Value on Accuracy

table 1 Table 1 .
Symptoms of Network Disorders

Table 3 .
Internet Access Limits

Table 4 .
Network Cable is Unplugged

Table 5 .
IP Address Conflict